As someone who is so tone deaf that she was asked to leave a music class because the teacher couldn’t believe anyone could be so bad at it, I never thought I could make music.
Decades later, I am sitting in my living room and making music thanks to MusicGen, the latest from Meta. One can use both text as well as existing melody to create something new.
Meta isn’t the first to try this. Sometime in January, it was announced that Google had “an impressive new AI system” called MusicLM that could generate music in any genre from text alone. But the company thought it was risky to put it out.
Image credit: Timmy Loen
Now Meta has gone another step ahead and made MusicGen open source. While I think it’s too early to declare who’ll come out the winner here, some say Meta’s platform seems to be more consistent when it comes to “creating higher quality audio.”
The model has reportedly been trained on 20,000 hours of music, which include 10,000 licensed songs and 390,000 instrumental tracks from stick media libraries.
The quality of songs being produced with MusicGen will undoubtedly keep getting better. However, we aren’t anywhere close to addressing some of the more ethical questions. These are concerns about artists’ rights, copyright issues, and ethical obligations, most of which have been relegated to the backburner.
In an era of deepfakes, I often find myself thinking about how cultural icons long dead and gone would react to a time where pretty much anything is possible, even deepfaking oneself because of a medical diagnosis, like in the case of the now-debunked Bruce Willis incident.
Now, Paul McCartney of The Beatles has announced that the “last” Beatles song will be released later this year featuring the voice of John Lennon.
While Lennon left the band long before his death, experts have been able to use AI to isolate the late musician’s voice in a way that it can be used for a brand new track.
I suspect this isn’t exactly what Lennon was talking about when he was encouraging others to imagine a world without a heaven or a hell, but I guess anything’s possible in a world where I can use AI to string a few tunes together by just using the right words.
This was published as a part of AI Odyssey, a section on generative AI developments featured in Tech in Asia’s emerging tech newsletter.
Delivered every Tuesday via email and through the Tech in Asia website, this free newsletter breaks down the biggest stories and trends in emerging tech. If you’re not a subscriber, get access by registering here.